Nondestructive Detection of Authenticity of Thai Jasmine Rice Using Multispectral Imaging

Table 2

Comparison of discrimination performance obtained with LS-SVM and BPNN methods with the spectral data and the combined spectral and morphological features data.

	LS-SVM		BPNN
	MS	Accuracy (%)	MS	Accuracy (%)
Calibration set
Thai jasmine rice (n = 150)
Spectral	4	97.3	0	100
Spectral + morphology	4	97.3	0	100
Jasmine sticky rice (n = 150)
Spectral	33	78	1	99.3
Spectral + morphology	34	77.3	0	100
Simiao rice (n = 150)
Spectral	14	90.6	0	100
Spectral + morphology	14	90.6	0	100
Northeast Wuchang rice (n = 150)
Spectral	4	97.3	0	100
Spectral + morphology	4	97.3	0	100
Total
Spectral	55	90.8	1	99.8
Spectral + morphology	56	90.6	0	100
Prediction set
Thai jasmine rice (n = 50)
Spectral	2	96	2	96
Spectral + morphology	1	98	2	96
Jasmine sticky rice (n = 50)
Spectral	18	64	6	88
Spectral + morphology	10	80	3	94
Simiao rice (n = 50)
Spectral	11	78	4	92
Spectral + morphology	12	76	7	86
Northeast Wuchang rice (n = 50)
Spectral	5	90	10	80
Spectral + morphology	9	82	4	92
Total
Spectral	36	82	22	89
Spectral + morphology	32	84	16	92

MS: misclassified samples.